meeting-assistant/docs/releases/v0.1.0-alpha.1.md

Meeting Assistant — First functional alpha

Version: 0.1.0a1 Tag: v0.1.0-alpha.1

This is the first functional end-to-end alpha of the local Meeting Assistant. It provides local meeting configuration and audio workflow, FFmpeg preparation and normalization, Whisper transcription, optional diarization, explicit speaker mapping, anonymous-speaker protocol generation, mapped-speaker regeneration, German and English meeting-language handling, glossary handling with YAML interchange, glossary provenance without transcript mutation, configurable protocol performance profiles, processing timings, generation history and provenance, and atomic publication of complete successful protocol generations. The GTM-Hub qualitative regression case is included.

The validated candidate revisions for this release were:

• Assistant: b50b87c32b3941310c719f459fd45faa87658b68
• Meeting Lab: d2e3636b949280402728308022c99bdee6ea1066

The immutable release revisions are the commits targeted by the paired v0.1.0-alpha.1 tags in the two repositories.

Accepted alpha limitations

• A local runtime, model, and container setup is required.
• Generated protocols require human review, especially responsibility and action attribution; speaker identity requires explicit confirmation.
• Diarization can be imperfect. The final native smoke environment did not contain pyannote.audio==4.0.7 and PyTorch, so real diarization and mapped speaker regeneration were not rerun there; automated tests cover these paths.
• Real German and English qwen3.8:27b protocol generation was validated.
• Context limits can cause attribution fallback or oversized-input rejection.
• Historical ignored Lab fixtures are required by some old experiment tests and are not automatically present in a completely fresh Lab worktree.
• Generation publication assumes a local POSIX filesystem; power-loss durability is not guaranteed.
• Visual redesign is post-alpha, and concise distribution-protocol rendering remains future work.